Optimum design of semi-rigid connections using metamodels

Considerable efforts were made in the past 15 years to develop strategies for the optimization of steel frames with semi-rigid connections, concentrating on the frames and not the connections, which were designed after the rest of the structure had been optimized. The analysis of semi-rigid connections requires the calculation of the moment-rotation curve (Mj-ϕ), which can be predicted using the Finite Element (FE) method. This is computationally expensive due to both the high number of degrees of freedom in the FE model and the nonlinear analysis required. In order to optimize such connections, a surrogate or metamodel of the FE model can be used. This paper puts forward a methodology for the optimal design of semi-rigid steel connections using metamodels generated with Kriging and Latin Hypercube, and optimized with the genetic algorithm method. This methodology was applied to two examples involving bolted extended end-plate connections, and was shown to work excellently at obtaining their optimal designs.

► Optimum design methodology of semi-rigid steel connections using metamodels.
► The Kriging Model and Latin Hypercube Sampling were used to build the metamodels.
► Two examples were optimized to demonstrate the methodology.
► The methodology was successfully implemented in Matlab.